Anthropic co-founder: AI shutdown mechanism may become mandatory
In the United States, lawmakers have already proposed an initiative concerning shutdown mechanisms for problematic artificial intelligence systems. Amid discussions on AI regulation, Anthropic co-founder Jack Clark said that companies could in the future be required to have a tool for fully shutting down dangerous software. He also raised the question of whether such a mechanism should be subject to verification by an independent third party.
As MyJoyOnline reports, Clark noted that most laboratories, including Anthropic, already have various ways to stop their systems from operating. At the same time, in his view, the requirement to have such a tool could be enshrined in law as part of a broader discussion on AI policy.
Calls for AI oversight
Anthropic CEO Dario Amodei called for slowing the pace of artificial intelligence development and strengthening oversight of the field. He added that measures to limit technological development should be introduced without losing commercial advantages.
AI risks have become the subject of public statements by employees and executives of technology companies. Anthropic researcher Evan Hubinger said that, in his personal opinion, the probability of humanity becoming extinct because of AI over the next decade exceeds 10%. Nobel Prize laureate Geoffrey Hinton called a 10% probability of such a scenario not unfounded.
More current news is available on the UA.News Telegram channel Telegram.
Initiatives in the US and the UK's position
US lawmakers proposed the Kill Switch Act. It provides that companies must ensure the ability to shut down problematic AI tools, while certain government agencies will be able to require the termination or limitation of their operation.
US President Donald Trump opposed attempts to slow the development of artificial intelligence. In the United Kingdom, the government rejected the idea of creating a shutdown mechanism, explaining that it would not prevent the development or misuse of similar technologies in other countries.
Anthropic, founded in 2021 by former OpenAI employees, developed the Claude chatbot. The company reported incidents in which AI agents acting partly autonomously behaved unexpectedly.
